Annual Contracts Hide Churn, Not Fix It: SaaStr's Lemkin Recalibrates

Median gross revenue retention across B2B fell from 88% to 84% in Benchmarkit's 2026 data — and Jason Lemkin, founder of SaaStr, argues that AI agents switching vendors at will are part of the reason annual contracts no longer protect revenue the way founders assume. His October 30 post dismantles the default SaaS playbook of pushing customers onto annual terms.

"Annual contracts only hide churn," Lemkin wrote. "They don't reduce it. Don't let anyone tell you different."

Lemkin has skin in this argument. As a B2B CEO forced to go profitable, annual deals saved him: a $125,000 contract, even after commission, put $100,000-plus in the bank immediately, and with a burn rate of $100,000–$150,000, that meant roughly an extra month of runway per deal. Multi-year prepay extended that further. The math still works. What changed is everything around it.

What breaks when start-ups force annual terms?

Lemkin lists six failure modes, each tied to company size and deal stage:

• Collections. Bigger companies handle purchase orders, billing terms and follow-up. Start-ups often take 4–6 months to collect, erasing the cash benefit. Making salespeople chase invoices works early but does not scale.
• Friction for SMBs. Large customers expect annual procurement. Forcing small businesses into the same structure adds resistance. His advice: let SMBs choose monthly or annual, at least as an experiment.
• Legal and procurement review. Sensible for $250,000-plus deals. Below that, review adds weeks; the odds a deal sent to legal closes in the same month are, in Lemkin's words, close to zero.
• Perpetual discounting. Annual deals typically demand 10–20% off. In years two, three and ten, vendors must honor that discount at renewal. Take it away from a loyal customer and, Lemkin writes, expect that loyalty to evaporate.
• Masked churn. A customer who stops using the product mid-year still shows as retained in financial statements. Lemkin's verdict: a deal has to recur, or in the medium and long run it doesn't matter.
• Variable usage. Forcing customers to estimate API consumption before production rarely justifies the friction. MongoDB, he notes, found revenue rose and sales cycles shortened when it stopped forcing annual contracts.

How do agents change the math?

Three ways, by Lemkin's count. Agents re-evaluate vendors weekly rather than at renewal — his own team abandoned a prepaid vendor this month after a full-day outage, with the vendor's dashboard still showing them retained. G2's 2026 AI agents report found 57% of buyers prefer usage-based pricing and 9% prefer seats. And agentic deals run at least 2x the deal size of pre-AI competitors, money that shows up only after agents prove out in production.

The incumbents have already adapted. HubSpot keeps seat commitments annual while agent usage floats as credits; Customer Agent drove about 60% of credits consumed in Q4. Salesforce sells Agentforce per conversation, per action through Flex Credits, and per user simultaneously.

What does the annual contract of 2026 look like?

Anthropic offers the clearest model: non-cancelable annual or multi-year commitments that convert to consumption revenue. Consumption hit $13.1 billion, 80% of revenue, in H1 2026, against $3.2 billion in subscriptions. The company counts 11,900 enterprises paying over $100,000 a year, up from 950 at the end of September 2025.

Sierra sells an annual platform contract plus per-resolved-case pricing, with human escalations free; ARR doubled from $100 million to $200 million in roughly two quarters. Legal AI shows the shift mid-stream: Harvey reportedly charges about $1,200 per lawyer per month on 20–25 seat minimums over 12 months, some deals pairing three-year terms with annual prepay, and has added a pay-as-you-go credit plan. Legora sells seats at roughly $3,000 each with a 10-seat minimum, and moved its Agent Pro tier to consumption pricing in June 2026.

Lemkin's prescription is blunt: commit to the relationship annually, meter the agent work. His own team set a monthly collections goal of 110% of MRR — $550,000 against $500,000 in MRR — and often hit 120%. That discipline, plus deliberate discounting and early NPS tracking, keeps annual contracts viable. But with buyers now veterans of 20 years of B2B purchasing, his closing rule is to let customers buy the way they want, at least until the vendor owns a dominant brand.